@ariacompute/engine-rn

import { AriaEngine } from '@ariacompute/engine-rn';
const eng = new AriaEngine('/path/to/bundle');
console.log(await eng.complete([{ role: 'user', content: 'hi' }]));

npm publish name: @ariacompute/engine-rn. Device-farm CI: .github/workflows/bindings-mobile.yml.

AriaEngine.open installs libaria_ffi into ~/.ariacompute/lib/ when it is not already on ARIA_FFI_LIB or in that cache (same Releases asset as aria-engine upgrade).

Auto-download by model name

AriaEngine.open(modelRef, { token, site }) accepts a local bundle path or an Aria model name (e.g. gemma-4-e2b-it_q4). A value containing / or already on disk is loaded directly; otherwise the SDK downloads it from the regional public hub (same as aria-engine download: .com → Hugging Face, .cn → ModelScope; site defaults to https://ariacompute.com) into ~/.ariacompute/models/{model} and then loads it. Dashboard is not used. A Dashboard sk- / sk-bf- token is ignored for hub auth. Token is optional for public models. Gated files: pass hfToken / modelscopeApiToken (same as aria-engine setup); if omitted, reads ~/.ariacompute/engine.yml. Instance setup is in-memory only (does not write that file). A valid cached bundle is reused without re-downloading.

const eng = await AriaEngine.open("gemma-4-e2b-it_q4");
const gated = new AriaEngine();
gated.setup({ hf_token: "hf_..." });
await gated.open("gemma-4-e2b-it_q4");